CANFIS PSO-Based Modeling of the Crazyflie Quadrotor

Understanding the dynamics of QUAV is essential for its effective use in real-world applications. This paper introduces the CANFIS architecture, a hybrid ML technique, to estimate the nonlinear dynamics of the QUAV. To ensure optimal performance and efficiency, the PSO technique was used in assigning the proposed model’s hyperparameters along with Recursive Least Squares (RLS) and Error Back-Propagation (EBP) to set the model’s parameters. The model is developed to capture both the translational and rotational dynamics of the CF drone. The model’s results demonstrate how the model’s outputs align with experimental data. In addition, a complete statistical analysis was performed to assess and validate the model’s accuracy. For translational dynamics, the model achieved MSE values of [Formula: see text] [Formula: see text] during the training phase and [Formula: see text] [Formula: see text] during the testing phase, with a VAF exceeding [Formula: see text] in both phases. For rotational dynamics, the MSE values were [Formula: see text] [Formula: see text] and [Formula: see text] [Formula: see text], with a VAF greater than [Formula: see text] in both phases. When benchmarked with existing models, this approach highlights the ability of the proposed framework to accurately capture the nonlinear dynamics of QUAVs, demonstrating remarkable accuracy and generalizability.